What it is. A long-arc strategy campaign built on four verbs — eXplore, eXpand, eXploit, eXterminate — where dozens of shallow systems (tech, economy, diplomacy, war) compound into an empire, and the drama is whether a snowballing lead stays interesting to the end.
Player fantasy / why it’s fun. You start with one settler in fog and end ruling a continent you built by hand. The pull is authored ownership: every border, tech, and war is a decision you can trace, and “one more turn” is the sound of a compounding plan paying off.
Pillars (exactly 3)
- Meaningful expansion. Each new claim — a city, a border, a colony — must open genuinely new options, not just add a bigger number. Land you took should let you do something you couldn’t before.
- Layered compounding systems. No single deep system; many shallow ones (economy, tech, culture, war, diplomacy) that multiply. Depth lives in the interactions, so the same board plays differently every campaign. See
pattern-emergence. - Snowball braced by comeback. The lead must accelerate (that’s the fantasy) and stay contestable. Build the
pattern-feedback-loopsthat reward a lead, then the negative loops that keep the loser in the game. Lose this brace and you getantipattern-solved-metagame.
The loop stack
| Scale | Beat |
|---|---|
| Moment | Place one tile / assign one worker / pick one build — a single small allocation. |
| Encounter | Resolve a war, a tech race, or a wonder rush against a specific rival over several turns. |
| Session | Take an empire through an era: settle, tech up, hit a power spike, project it outward. |
| Meta | The full arc from one settler to a win condition — and the next start, wanting to try a different opening. |
The dangerous scale is Meta: 4X campaigns run long, and the back half is where the lead is often already decided. Design the endgame first, not last.
Essential systems
system-tech-tree— the spine of eXploit; gates new verbs and power spikes across the arc. Keep branches that change how you play, not just +stat leaves (antipattern-false-depth).system-economyandsystem-resource-loops— the yields (food, production, science, gold) whose ratios are the strategy.system-faction-asymmetry— leaders/civs that start the four X’s in different postures; the reason to replay.system-progression— eras/ages that re-baseline power so an early lead doesn’t run away untouched.system-difficulty-and-dda— AI handicaps and catch-up bands; the honest lever against runaway.system-procgen-design— map generation is the opening variety engine; a fresh continent is a fresh puzzle.system-enemy-ai— opponents must expand and fight competently, or the mid-game goes hollow. This is 4X’s hardest build.system-map-and-navigationandsystem-inventory-and-ui— a legible board and a readable stack of menus; a 4X lives and dies onpattern-readabilityat density.
Content & difficulty model
- Content is combinatorial, not authored. The map, the civ, the neighbors, and the tech order recombine into the “content.” Your job is a generator with enough orthogonal axes that no two starts feel the same. See
system-procgen-design. - Difficulty is asymmetry + information, then handicap. First tune what the AI does; only then add resource multipliers. A pure-bonus AI reads as a cheat, not a rival.
- Ramp is intra-campaign. The threat should escalate as you snowball — later eras, stronger rivals, unlocked win-races — so power spikes meet fresh pressure, per
pattern-pacing-and-tension. - Win conditions are the difficulty knob for the endgame. Multiple, racing win paths (domination, science, culture, score) keep a losing-on-war player alive on another axis. This is your primary defense against the solved lead.
Signature-mechanic seeds
- 4X but one city, and expansion is vertical (constraint). No new settlements; you eXpand inward — districts, tiers, and specializations stacked on a single capital. Every “conquest” is a building slot fought over. Turns stay fast, the map shrinks, the tech and economy layers carry the whole game. Kills turn-time bloat and the tall-vs-wide non-choice at once.
- 4X but you play the map, not a faction (perspective). You are the continent — climate, resources, disasters, migration — nudging the AI civilizations that settle you. eXplore becomes revealing your own body; eXterminate becomes deciding which empire your volcanoes and droughts favor. Turns the snowball inside-out: the drama is whose lead you allow.
- 4X but the tech tree is shared and public (constraint). One global tree; researching a leaf locks it or discounts it for whoever finishes second. Every science pick is a race and a denial. Turns
system-tech-treeinto a live auction and makes the mid-game contestable by construction. - 4X but eXterminate is banned (constraint). No unit can destroy another empire; you win only by expansion, culture, or economic vassalage. Removes the dominant runaway path and forces the other three X’s to carry tension — a cold-war 4X. Pairs with dense
system-economypressure. - 4X but the era resets the board (twist). Each age wipes accumulated units and re-rolls terrain around fixed capitals, carrying only tech and culture forward. A late lead is a soft lead, not a locked one; the comeback brace is baked into the calendar.
Common pitfalls
antipattern-solved-metagame— the boring solved-lead endgame: once ahead, the winner just executes. The genre’s signature failure. Fight it with racing win conditions, era re-baselines, and negative feedback loops.antipattern-feature-soup— piling on systems (religion + espionage + trade routes + …) that don’t interact. Every new system must multiply an existing one or cut it.antipattern-decision-paralysis— turn-time bloat: too many equal-weight micro-choices per turn. Automate the trivial (worker allocation, repeats), surface only the load-bearing decisions.antipattern-false-depth— a tech tree that’s all +10% leaves. If a node doesn’t change how you play, it’s a stat, not a choice; seesystem-tech-tree.antipattern-fake-choice— win conditions where one path dominates every start, so “your choice of victory” is really one path with flavor.antipattern-content-desert— the late game where the map is settled, the tree is done, and there’s nothing new to reach for. Reserve unlocks and pressure for the back half.
Anchors
anchor-civilization— the reference for the four X’s, the era arc, “one more turn,” and multiple win conditions. Study its ratio of shallow systems and how it tries (and sometimes fails) to keep the lead contestable.
For the neighbor genres: real-time expansion is genre-rts (see anchor-starcraft, anchor-age-of-empires); the combat-only zoom-in of eXterminate is genre-tactics (anchor-into-the-breach); pure-compounding-numbers is genre-incremental; a board-first, systems-light cousin is genre-abstract-strategy. For the feel of a lead compounding, borrow from genre-city-builder (anchor-frostpunk) and anchor-loop-hero.
Verify
Determinism is non-negotiable: a 4X is a huge state machine over hundreds of turns — all randomness through a deterministic RNG, ordered iteration for every AI and yield pass, so a seed replays identically. Prove the endgame isn’t solved (a trailing AI can still win on some axis) and that turn resolution stays readable at scale. Gates in docs/VERIFICATION.md; design the fun of the snowball-vs-comeback tension per pattern-feedback-loops and pattern-pacing-and-tension.